Bruce Momjian wrote: > Would some people please run the attached test procedure and report back > the results? I basically need to know the patch is an improvement on > more platforms than just my own. Thanks
Debian Sarge/AMD64 Kernel 2.6.16.16 (all tests done multiple times with variation of less then 10%): -HEAD: real 0m0.486s user 0m0.064s sys 0m0.048s -HEAD with 100000 "SELECT 1;" queries: real 0m4.763s user 0m0.896s sys 0m1.232s -HEAD + stats: real 0m0.720s user 0m0.128s sys 0m0.096s -HEAD + stats (100k): real 0m7.204s user 0m1.504s sys 0m1.028s -HEAD + stats + patch: there is something weird going on here - I get either runtimes like: real 0m0.729s user 0m0.092s sys 0m0.100s and occasionally: real 0m3.926s user 0m0.144s sys 0m0.140s (always ~0,7 vs ~4 seconds - same variation as Qingqing Zhou seems to see) -HEAD + stats + patch(100k): similiar variation with: real 0m7.955s user 0m1.124s sys 0m1.164s and real 0m11.836s user 0m1.368s sys 0m1.156s (ie 7-8 seconds vs 11-12 seconds) looks like this patch is actually a loss on that box. Stefan ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match